European Harmonised Standards

The AI Act establishes that the compliance of AI systems – especially those classified as high-risk systems – must be demonstrated based on harmonised technical standards developed by recognised European standardisation bodies, namely CEN (European Committee for Standardisation) and CENELEC (European Committee for Electrotechnical Standardisation).

These standards will serve as a technical reference for economic operators – such as suppliers, integrators, and users – enabling them to adapt their processes and products to the regulation’s requirements. The European Commission has already issued a standardisation request with a set of priorities to be developed by 2026, in coordination with the AI Office and Member States.

The standards will cover areas such as:

• Quality and representativeness of data used for model training;
• Methods for risk assessment and mitigation;
• System robustness and cybersecurity;
• Algorithm transparency and explainability;
• Human oversight and controllability;
• Registration requirements and technical documentation.

Although their adoption is voluntary, compliance with these standards provides a presumption of conformity with the regulation’s legal requirements. This principle allows companies to reduce administrative burdens and accelerate the assessment and market deployment of their AI systems.